HAL Id: hal-01779813
https://hal.archives-ouvertes.fr/hal-01779813
Submitted on 27 Apr 2018
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Cosmological constraints from galaxy clustering in the
presence of massive neutrinos
C Zennaro, J Dossett, Julien Bel, C. Carbone, L Guzzo
To cite this version:
C Zennaro, J Dossett, Julien Bel, C. Carbone, L Guzzo. Cosmological constraints from galaxy
clus-tering in the presence of massive neutrinos. Monthly Notices of the Royal Astronomical Society,
Oxford University Press (OUP): Policy P - Oxford Open Option A, 2018, 477 (1), pp.491 - 506.
�10.1093/mnras/sty670�. �hal-01779813�
Cosmological constraints from galaxy clustering in the
presence of massive neutrinos
M. Zennaro
1,3, J. Bel
2,3, J. Dossett
3, C. Carbone
1,3,4, L. Guzzo
1,31Universit`a degli studi di Milano - Dipartimento di Fisica, via Celoria, 16, 20133 Milano, Italy 2Aix Marseille Univ, Universit´e de Toulon, CNRS, CPT, Marseille, France
3INAF - Osservatorio Astronomico di Brera, Via Brera, 28, 20121 Milano, Italy 4INFN - Sezione di Milano, via Celoria, 16, 20133 Milano, Italy
Received date / Accepted date
ABSTRACT
The clustering ratio is defined as the ratio between the correlation function and the variance of the smoothed overdensity field. In ΛCDM cosmologies not accounting for massive neutrinos, it has already been proved to be independent from bias and redshift space distortions on a range of linear scales. It therefore allows for a direct comparison of measurements (from galaxies in redshift space) to predictions (for matter in real space). In this paper we first extend the applicability of such properties of the cluster-ing ratio to cosmologies that include massive neutrinos, by performcluster-ing tests against simulated data. We then investigate the constraining power of the clustering ratio when cosmological parameters such as the total neutrino mass Mν and the equation
of state of dark energy w are left free. We analyse the joint posterior distribution of the parameters that must satisfy, at the same time, the measurements of the galaxy clustering ratio in the SDSS DR12, and the angular power spectrum of temperature and polarization anisotropies of the CMB measured by the Planck satellite. We find the clustering ratio to be very sensitive to the CDM density parameter, but not very much so to the total neutrino mass. Lastly, we forecast the constraining power the clustering ratio will achieve with forthcoming surveys, predicting the amplitude of its errors in a Euclid-like galaxy survey. In this case, we find it is expected to improve the constraint at 95% level on the CDM density by 40% and on the total neutrino mass by 14%.
Key words: bias, clustering, neutrinos
1 INTRODUCTION
Present-time as well as forthcoming galaxy surveys, while on the one hand will allow us to reach unprecedented precision on the measurement of the galaxy clustering in the universe, on the other hand will challenge us to produce more accur-ate and reliable predictions. The effect of massive neutrinos on the clustering properties of galaxies, that in the past has been either neglected or considered as a nuisance para-meter, is nowadays regarded as one of the key points to be included in the cosmological model in order for it to reach the required accuracy. At the same time, while allowing for more realistic predictions of cosmological observables, this process also helps in shedding light on some open issues of fundamental physics, such as the neutrino total mass or the hierarchy of their mass splitting.
From the experiments measuring neutrino flavour oscil-lations, particle physics has been able to draw a constraint on the mass splitting of the massive eigenstates of
neut-rinos, and set a lower bound to the total neutrino mass,
Mν = P mν,i & 0.06 eV at 95% level (Gonzalez-Garcia
et al. 2012, 2014; Forero et al. 2014; Esteban et al. 2017). On the other hand, the absolute scale of magnitude of neutrino masses is still an open issue. Beta decay experi-ments such as the ones carried out in Mainz and Troitsk have set as an upper limit at 95% level on the electron neutrino
mass of m(νe) < 2.2 eV (Kraus et al. 2005). While future
experiments like Katrin prospect much higher sensitivities, of order 0.2 eV (Bonn et al. 2011), present day cosmology can already intervene in the debate about neutrino mass.
Since neutrinos are light and weakly interacting, they decouple from the background when still relativistic. There-fore, even at late times they are characterised by large ran-dom velocities that prevent them from clustering on small scales. As a consequence, neutrinos introduce a character-istic scale-dependent and redshift-dependent suppression of the clustering, whose amplitude depends on the value of
their mass. In fact, the presence of massive neutrinos influ-ences the evolution of matter overdensities in the universe, depending on their mass. Last years have seen a large num-ber of works extensively studying the interplay between cos-mology and neutrino physics (see, for instance, Lesgourgues & Pastor 2006, 2012, 2014, and references therein).
As we aim to describe the clustering of galaxies in cos-mologies with massive neutrinos, we have to cope with the description of the galaxy-matter bias. As a matter of fact, galaxies do not directly probe the matter distribution in the universe, being in fact a discrete sampling of its highest dens-ity peaks. We choose to describe galaxy clustering through a recently introduced observable that, on sufficiently large scales, does not depend on the galaxy-matter bias, the clus-tering ratio (Bel & Marinoni 2014).
In standard ΛCDM cosmologies, this observable has already been proved to be a reliable cosmological probe for constraining cosmological parameters, being particularly sensitive to the amount of matter in the universe, as shown in Bel et al. (2014).
In this work we aim at studying how the clustering prop-erties of galaxies are modified by the presence of neutrinos, and in particular we want to extend the clustering ratio ap-proach to cosmologies including massive neutrinos. By prov-ing that this observable maintains its properties, we want to exploit it to constrain the total neutrino mass.
This paper will be organized as follows. In Sec. 2 we will introduce the statistical observable we are going to use, the clustering ratio, and its properties. We will show why this observable can be considered unaffected either by the galaxy-matter bias on linear scales and redshift-space dis-tortions, and we will introduce its estimators.
In Sec. 3 we will describe the effects of massive neutri-nos on the matter and galaxy clustering. We will introduce the DEMNUni simulations, the set of cosmological simula-tions we use to test the properties of the clustering ratio in a cosmology with massive neutrinos. Finally we will show that the properties of the clustering ratio hold as well in cosmo-logies that include massive neutrinos, in particular confirm-ing the independence of the clusterconfirm-ing ratio from bias and redshift-space distortions on linear scales in the DEMNUni simulations.
Sec. 4 is devoted to presenting our results. We use a measurements of the clustering ratio in the Sloan Digital Sky Survey Data Release 7 and 12 to draw a constraint on the total neutrino mass and on the equation of state of dark energy. In particular we study the joint posterior
distribution of the parameters of the model, including Mν
and w, obtained from the clustering ratio measurement and the latest cosmic microwave temperature and polarization anisotropy data from the Planck satellite.
2 CLUSTERING RATIO
In order to describe the statistical properties of the matter distribution in the universe, we introduce the overdensity field
δ(x, t) = ρ(x, t)
¯
ρ(t) − 1, (1)
where ρ(x, t) is the value of the matter density at each
spa-tial position, while ¯ρ(t) represents the mean density of the
universe.
This is assumed to be a random field with null mean. Information on the distribution must therefore be sought in
its higher order statistics, such as the variance σ2 = hδ2(x)i
c
and the 2-point autocorrelation function ξ(r) = hδ(x)δ(x +
r)icof the field. Here h·icdenotes the cumulant moment, or
connected expectation value (Fry 1984).
In this work we will always consider the matter distribu-tion smoothed on a certain scale R by evaluating the density contrast in spherical cells, i.e.
δR(x) = Z δ(x0)W |x − x 0 | R d3x0, (2)
where W is the spherical top-hat window function. As a consequence, the variance and correlation function will be
smoothed on the same scale, and will be denoted σR2 and
ξR(r).
An equivalent description of the statistical properties of the matter field can be obtained in Fourier space in terms of the matter power spectrum. Starting from the Fourier transform of the matter overdensity field,
ˆ δ(k) = Z d3x (2π)3e −ik·x δ(x) , (3)
the matter density power spectrum is defined according to
hˆδ(k1)ˆδ(k2)i = δD(k1+ k2)P (k1) , (4)
while the adimensional power spectrum can be written as
∆2(k) = 4πP (k)k3. The variance and correlation function
of the matter field are linked to its power spectrum, repres-enting in fact different ways of filtering it. The variance is the integral over all the modes, modulated by the Fourier
transform of the filtering function ˆW ,
σR2 =
Z ∞
0
∆2(k) ˆW2(kR) d ln k, (5)
and the correlation function is, in addition, modulated by
the zero-th order spherical Bessel function j0(x) = sin(x)/x,
ξR(r) =
Z ∞
0
∆2(k) ˆW2(kR)j0(kr) d ln k. (6)
The explicit expression of ˆW (kR) is
ˆ
W (kR) = 3
kRj1(kR) = 3
sin(kR) − kR cos(kR)
(kR)3 . (7)
However, in practice, we are not able to directly access the matter power spectrum. The reason is that the galaxies we observe do not directly probe the distribution of mat-ter in the universe. In fact, they represent a discrete biased sampling of the underlying matter density field and the bias-ing function is, a priori, not known. A way to overcome this problem is to refine the independent measurements of the bias function through weak lensing surveys. Otherwise, one can parametrise the bias adding additional nuisance para-meters to the model and, consequently, marginalize over them.
On the other hand, a completely different approach is to seek new statistical observables, which can be considered
to be unbiased by construction. This is the path followed by Bel & Marinoni (2014) by introducing the clustering ratio,
ηR(r) ≡
ξR(r)
σ2
R
, (8)
that is the ratio of the correlation function over the variance of the smoothed field.
We assume that the relation between the matter density contrast and the galaxy (or halo) density field is a local and deterministic mapping, which is enough regular to allow a Taylor expansion (Fry & Gaztanaga 1993) as
δg,R= F (δR) ' N X i=1 bi i!δ i R. (9)
Moreover we assume the growth of fluctuations to occur hier-archically (see Bernardeau et al. 2002; Bel & Marinoni 2012), so that each higher order cumulant moment can be expressed according to powers of the variance and 2-point correlation funtion, hδn Ric= Snσ 2(n−1) R , hδi,Rn δ m j,Ric= CnmξR(r)σR2(n+m−2), (10) the former applying to the 1-point statistics and the latter to the 2-point ones.
It has been shown by Bel & Marinoni (2012) that the bias function only modifies the clustering ratio of galaxies at the next order beyond the leading one and that it is not sensitive to third order bias
ηg,R(r) ' ηR(r) + 1 2c 2 2ηR(r)ξR(r) − (S3,R− C12,R)c2+ 1 2c 2 2 ξR(r) (11)
where c2 ≡ b2/b1. By choosing a sufficiently large
smooth-ing scale, the higher order contribution in Eq. (11) becomes negligible and we obtain
ηg,R(r) = ηR(r), (12)
meaning that in this case the clustering ratio of galaxies can be directly compared to the clustering ratio predicted for the matter distribution.
The local biasing model is not the best way of describ-ing the bias function between matter and haloes/galaxies (Mo & White 1996; Sheth & Lemson 1999; Somerville et al. 2001; Casas-Miranda et al. 2002). It can be improved by introducing a non-local component depending on the tidal field. However, it has been shown by Chan et al. (2012) and by Bel et al. (2015) that, when dealing with statistical quantities which are averaged over all possible orientations, then the non-local component is degenerate with the second
order bias c2, thus expression (11) remains valid and we do
not consider non-local bias in our analysis.
On linear scales, the clustering ratio is expected to be independent from redshift. Since, in a ΛCDM universe, we can write (normalizing all quantities to the present time
vari-ance on the scale r8= 8 h−1Mpc, σ82(z = 0)) the evolution
of the variance and the correlation function as
σ2R(z) = σ 2 8(z = 0)D 2 (z)FR ξR(r, z) = σ82(z = 0)D 2 (z)GR(r), (13)
where D(z) is the linear growth factor of matter density fluctuations and FR= R∞ 0 ∆ 2 (k) ˆW2(kR)d ln k R∞ 0 ∆ 2(k) ˆW2(kr 8)d ln k GR(r) = R∞ 0 ∆ 2 (k) ˆW2(kR)j0(kr)d ln k R∞ 0 ∆ 2(k) ˆW2(kr 8)d ln k ,
depend only on the shape of the power spectrum. Hence, the
clustering ratio ξR(r)/σR2 = GR(r)/FR, does not depend on
D(z), which cancels outs.
In practice, we include weak nonlinearities which intro-duce a small, but nevertheless detectable, redshift depend-ence.
Not only measures of the clustering of galaxies are biased with respect to predictions for the matter field, but they also are affected by the peculiar motion of galaxies. This motion introduces a spurious velocity component (along the line-of-sight) that distorts the redshift assigned to galax-ies. Since, for the clustering ratio, we are interested in large smoothing scales and separations, we can focus on the linear scales, where the only effect is due to the coherent motion of infall of galaxies towards the overdense regions in the uni-verse.
We can link the position of a galaxy (or dark matter halo) in real space to its apparent position in redshift-space. Let us denote r as the true comoving distance along the line-of-sight; in redshift space it becomes
s = r +vpk (1 + z)
H(z) eˆr, (14)
where vpkis the line-of-sight component of the peculiar
velo-city and ˆeris the line-of-sight versor. Considering the
Four-ier space decomposition of the density contrast, the relation linking its value in redshift space to the one in real space (Kaiser 1987) is
δs(k) = (1 + f µ2)δ(k), (15)
where quantities in redshift space are expressed with the superscript s and µ is the cosine of the angle between the wavemode k and the line-of-sight. Here f is the so called growth rate, defined as the logarithmic derivative of the growth factor of structures with respect to the scale factor, f ≡ d ln D/d ln a. Averaging over all angles ϑ, the variance and the correlation function in redshift space result modified by the same multiplicative factor
σs 2R = Kσ
2 R
ξRs(r) = KξR(r)
(16)
where K = 1 + 2f /3 + f2/5 is the Kaiser factor. As a
con-sequence, the clustering ratio is unaffected by redshift-space distortions on linear scales. This argument allows us to re-write the identity (12) as
ηsg,R(r) ≡ ηR(r), (17)
meaning that, by properly choosing the smoothing scale R and the correlation length r, measures of the clustering ratio from galaxies in redshift space can be directly compared to predictions for the clustering ratio of matter in real space.
2.1 Estimators
The clustering ratio can be estimated from count-in-cells, where, under the assumption of ergodicity, all ensemble av-erages become spatial avav-erages. We follow the counting pro-cess set up by Bel & Marinoni (2012), we define the discrete density contrast as
δN,i=
Ni
¯
N − 1, (18)
where Ni is the number of objects in the i -th cell and ¯N is
the mean number of objects per cell. The estimator of the variance is therefore ˆ σ2R= 1 p p X i=1 δi2 (19)
and the one of the correlation function is ˆ ξR(r) = 1 pq p X i=1 q X j=1 δiδj (20)
leading to the definition of the estimator of the clustering ratio as ˆ ηR(r) = ˆ ξR(r) ˆ σ2 R . (21)
Throughout this work we will often express the correlation length r as a multiple of the smoothing scale, i.e. r = n R.
Since we are dealing with a discrete counting process, the shot noise needs to be properly accounted for. We fol-low the approach of Bel & Marinoni (2012) and correct the estimator of the variance according to
ˆ σR2 = hδ 2 n(x)i − 1 ¯ N = 1 p p X i=1 δ2i − 1 ¯ N, (22)
where ¯N is the mean number of objects per cell. On the
other hand, the correlation function needs no correction, as long as the spheres do not overlap.
2.2 Effects of massive neutrinos
We introduce massive neutrinos as a subdominant dark mat-ter component. For simplicity, we consider three
degener-ate massive neutrinos, with total mass Mν =Pimν,i and
present-day neutrino energy density in units of the critical
density of the universe Ων,0h2= Mν/(93.14 eV). The
neut-rino fraction is usually expressed with respect to the total
matter as fν = Ων/Ωm. For a more complete treatment of
neutrinos in cosmology, we refer the reader to Lesgourgues & Pastor (2006, 2012, 2014).
Neutrinos of sub-eV mass, which seem to be the most likely candidates both from particle physics experiments and cosmology, decouple from the primeval plasma when the weak interaction rate drops below the expansion rate of the universe, at a time when the background temperat-ure is around T ' 1 MeV. This corresponds to a redshift
1 + zdec ∼ 109. Since the redshift of their non-relativistic
transition, obtained equating their rest-mass energy and their thermal energy, is given by
1 + znr' 1890
mν,i
1 eV, (23)
when neutrinos decouple, they are still relativistic. As a con-sequence, since the momentum distribution of any species is frozen at the time of decoupling, neutrino momenta keep following a Fermi-Dirac distribution even after their non-relativistic transition, and neutrinos end up being character-ised by a large velocity dispersion. An effective description of the evolution of neutrinos can be achieved employing a fluid approximation (Shoji & Komatsu 2010). In this
frame-work we can define a neutrino pressure, pν= wνρνc2,
com-puted integrating the momentum distribution. Such pres-sure is characterised by an effective adiabatic speed of sound (Blas et al. 2014)
cs,i= 134.423 (1 + z)
1eV
mν,i
km s−1, (24)
that represents the speed of propagation of neutrino dens-ity perturbations. Such speed of sound defines the min-imum scale under which neutrino perturbations cannot grow, called the free streaming scale. It corresponds to a wavenumber kF S(z) = 4πG¯ρa2 c2 s 1/2 = 3 2 H2Ω m(z) (1 + z)2c2 s 1/2 , (25) or a proper wavelength λF S= 2πa/kF S. (26)
At each redshift, neutrino density fluctuations of wavelength smaller than the free streaming scale are suppressed, their gravitational collapse being contrasted by the fluid pressure support. As a consequence, neutrinos do not cluster on small scales and remain more diffuse compared to the cold matter component.
Neutrino free streaming does not only affect the evolu-tion of neutrino perturbaevolu-tions, in fact it affects the evoluevolu-tion of all matter density fluctuations. We can model the growth of matter fluctuations employing a two-fluid approach (Blas et al. 2014; Zennaro et al. 2017). In this case, the solution of the equations of growth for the neutrino and cold matter fluids are coupled,
¨ δcb+ H ˙δcb− 3 2H 2Ω m{fνδν+ (1 − fν)δcb} = 0 ¨ δν+ H ˙δν− 3 2H 2Ω m fν− k2 k2 F S δν+ (1 − fν)δcb = 0, (27) where derivatives are taken with respect to conformal time, dτ = dt/a, and both the Hubble function and the matter density parameter are functions of time, H = H(τ ) and
Ωm= Ωm(τ ).
The coupling of these equations requires the evolution of the CDM density contrast to be scale dependent, unlike in standard ΛCDM cosmologies. We therefore expect to find an observable suppression even in the CDM+baryon power spectrum, starting from the mode corresponding to the size of the free-streaming scale at the time of the neutrino
non-relativistic transition, knr= kF S(znr), and affecting all the
ΛCDM NU0.17 NU0.30 NU0.53 Mν[eV] 0 0.17 0.30 0.53
Ωc 0.27 0.2659 0.2628 0.2573
σ8,cc 0.846 0.813 0.786 0.740
Table 1. The cosmological parameters that vary among the 4 DEMNUni simulations considered in the present work, depending on the assumed neutrino total mass.
3 CLUSTERING RATIO WITH MASSIVE
NEUTRINOS
In order to investigate the behaviour of the clustering ratio in cosmologies with massive neutrinos, i.e. whether it main-tains all the properties described in Sec. 2, we analyse the cosmological simulations “Dark Energy and Massive Neut-rino Universe” (DEMNUni), presented in Castorina et al. (2015) and Carbone et al. (2016).
These simulations have been performed using the Gadget-III code by Viel et al. (2010) based on the Gadget simulation suite (Springel et al. 2001; Springel 2005). This version includes three active neutrinos as an additional
particle species1.
The DEMNUni project comprises two set of simula-tions. The first one, which is the one considered in the present work, includes 4 simulations, each implementing a different neutrino mass. Besides the reference ΛCDM
simu-lation, which has Mν= 0 eV, the other ones are
character-ised by Mν= {0.17, 0.30, 0.53} eV. The second set includes
10 simulations, exploring different combinations of neutrino masses and dynamical dark energy parameters.
All simulations share the same Planck-like cosmology,
with Hubble parameter H0 = 67 km s−1 Mpc−1, baryon
density parameter Ωb = 0.05, primordial spectral index
ns = 0.96, primordial amplitude of scalar perturbations
As = 2.1265 × 109 (at a pivotal scale kp = 0.05 Mpc−1)
and optical depth at the time of recombination τ = 0.0925.
The density parameter of the cold dark matter, Ωcdm, is
ad-justed in each simulation, depending on the neutrino mass, so that all simulations share the same total matter density
parameter Ωm = 0.32, see Tab. 1. Each simulation follows
the evolution of 20483 CDM particles and, when present,
20483 neutrino particles, in a comoving cube of 2 h−1
Gpc
side. The mass of the CDM particle is ∼ 8 × 1010h−1M ,
and changes slightly depending on the value of Ωcdm. All
simulations start at an initial redshift zin = 99 and reach
z = 0 with 62 comoving outputs at different redshifts. In this work we focus on the snapshots at redshift z = 0.48551 and z = 1.05352.
Dark matter haloes have been identified through a Friend-of-Friends (FoF) algorithm with linking length b = 0.2 and setting the minimum number of particles needed to form a halo to 32. Thus, the least massive haloes have mass
of about 2.6 × 1012h−1
M . In order to check the stability
of our results regarding the choice of the definition of a halo
1 The simulations do not account for an effective neutrino number
Neff > 3, as possible neutrino isocurvature perturbations which
could produce larger Neff (therefore affecting galaxy and CMB
statistics Carbone et al. 2011) are currently excluded by present data (see, eg, Di Valentino & Melchiorri 2014)
Bin Mass range [1012h−1M ] 0 0.58 6 M < 1.16 1 1.16 6 M < 2.32 2 2.32 6 M < 3.28 3 3.28 6 M < 4.64 4 4.64 6 M < 6.55 5 6.55 6 M < 9.26 6 9.26 6 M < 30 7 30 6 M < 100 8 M > 100
Table 2. Subdivision of the halo catalogues in mass bins.
we also have access to halo catalogues where haloes have been identified using spherical over-densities. For the pur-pose of the present work we constructed halo catalogues in redshift space by modifying the positions along the z direc-tion according to the projected velocity (properly converted in length) in that direction (see Eq. 14).
Regarding error estimation, being these simulations very large, a jackknife method has been implemented by subdividing the box in 64 sub-cubes. The standard error on
the measured value of ηR(r) is then taken to be the
disper-sion obtained from the jackknife process
ση2R= Nj− 1 Nj Nj X i=1 [ηR,i(r) − ¯ηR(r)]2, (28)
where Njis the number of jackknife resamplings, in our case
Nj= 64.
In the following, we first check the reliability of the clustering ratio in the presence of massive neutrinos. In particular we are interested in proving that the identity
ηzR,g(r) ≡ ηR(r) still holds. To this end, we must prove that
the clustering ratio at the scales of interest does not depend on the galaxy-matter bias (so that we can compare predic-tions for matter and measures from galaxies) and that it is not affected by redshift-space distortions (to be safe when comparing the real space predictions to measures obtained in a galaxy redshift survey).
3.1 Bias sensitivity
In order to test the independence of the clustering ratio from the bias on linear scales, we divide the dark matter haloes in 9 mass bins, reported in Tab. 2. The various halo popula-tions evolve in a different way, therefore they present differ-ent biasing functions with respect to the dark matter field.
Thus, we will use the linear bias bL to characterise each
halo sample. In Tab. 3 we show how both the FoFs and the spherical overdensities from the simulations populate these mass bins in the simulations. Due to the minimum number of particle required to identify a halo, the first two mass bins do not contain any. On the other hand, the only mass limit to the spherical over-densities is given by the mass resolution of the simulation, hence all the mass bins are populated.
In Fig. 1 we show the estimated correlation functions
of each FoF sample in the two extreme cases of Mν= 0 eV
and Mν= 0.53 eV (the same holds for the SOs as well).
We also represent the corresponding correlation func-tion of the dark matter field, which is used to estimate the
bin 0 bin 1 bin 2 bin 3 bin 4 bin 5 bin 6 bin 7 bin 8 FoF z = 0.48551 Mν= 0.00 eV 0 0 2902221 3509393 2402274 1708375 2878557 758008 145410 Mν= 0.17 eV 0 0 3152025 3178910 2430866 1667315 2712374 690356 122241 Mν= 0.30 eV 0 0 3116471 3269324 2263001 1642694 2589654 634844 104539 Mν= 0.53 eV 0 0 3273517 3026718 2144285 1501769 2334332 532432 76127 z = 1.05352 Mν= 0.00 eV 0 0 2571902 3039264 2003119 1358767 2044706 389064 38299 Mν= 0.17 eV 0 0 2713196 2674546 1958721 1277479 1836860 328973 28852 Mν= 0.30 eV 0 0 2620295 2674912 1767015 1218810 1679768 283298 22244 Mν= 0.53 eV 0 0 2619539 2335248 1570202 1036782 1386909 206588 13059 SO z = 0.48551 Mν= 0.00 eV 453415 2973241 2783664 2361431 1669494 1210036 2064503 527389 88178 Mν= 0.17 eV 471602 2993986 2904400 2118437 1668026 1166042 1917029 470522 72097 Mν= 0.30 eV 486007 2997527 2842491 2159522 1559049 1112575 1806821 424202 60046 Mν= 0.53 eV 508713 3259563 2582711 1956593 1424500 1014066 1587591 343288 41497 z = 1.05352 Mν= 0.00 eV 363148 2449897 2433901 2030689 1373779 952365 1446087 263959 22482 Mν= 0.17 eV 359806 2376103 2479355 1766898 1330746 885172 1280841 218832 16462 Mν= 0.30 eV 354938 2304946 2374992 1750620 1206312 819053 1157522 184575 12287 Mν= 0.53 eV 341645 2359735 2069025 1500656 1039892 694588 932357 130319 6834
Table 3. Population of the 9 mass bins for the Friend-of-Friends (FoF) and spherical overdenities with respect to the critical density (SO) at redshift z = 0.48551 and z = 1.05352.
linear bias bLcharacterizing each halo sample:
bL≡
s
ξRF oF(nR)
ξR(nR)
. (29)
We find that our cut in mass does indeed correspond to different tracers, with higher bias for higher-mass objects. However, such different halo populations still show a con-stant bias with respect to scale, which allows us to fit the measured bias in Fig. 1 with flat lines.
The independence of the FoF-matter bias from scale is confirmed also in the massive neutrino case (Fig. 1, right). In particular, we note here that the bias is generally higher when considering massive neutrinos. This is due to the fact that, since they smooth the matter distribution, neutrinos make haloes of a given mass rarer then in a standard ΛCDM cosmology.
Secondly we compute the clustering ratio for both the FoFs and the spherical overdensities at redshift z = 0.48551 and z = 1.05352. In each of these cases, we analyse the ΛCDM simulation, which does not include
massive neutrinos, and the ΛCDMν simulations with Mν=
{0.17, 0.30, 0.53} eV. In the two plots in Fig. 2 we show the
clustering ratio for fixed smoothing radius R = 16 h−1Mpc
and correlation length r = 2R in the same mass bins shown in Tab. 2 for the FoFs and spherical overdensities respect-ively. Points are measurements in the simulations, while lines are the predictions obtained from the matter power spec-trum. The ratio between measures and prediction is in the bottom panel. At z = 0.48551, the measured clustering ratio
in the bins with masses < 30.43h−1M agrees with the
pre-dictions for the matter at 3% level. Below 100 h−1M the
agreement is within 5%. For objects with mass greater than this, we observe a more scattered trend. We blame a lower statistical robustness, due to fewer objects falling in these mass bins. In any case, we do not observe any peculiar de-pendence of the clustering ratio on the mass of the objects, confirming up to a few percent accuracy its independence on the bias at these scales.
Conversely, at z = 1.055 we do see a dependence of the FoF clustering ratio on the mass. The reason is likely to
be a lack of mass resolution in the simulations, as moving towards higher redshifts not enough haloes have formed in the high-mass end of the mass function. For this reason, we expect this situation to be even more sever in the cases with masses neutrinos, which is confirmed in the right power of Fig. 2.
As we find similar results for both the FoFs and the spherical overdensities, we conclude that such results do not depend on the tracer we choose to observe.
Finally, we claim that the clustering ratio is insensitive to the bias on linear scales in a cosmology including massive neutrinos irrespective of either the mass of the tracer or the nature of the tracer itself or the total mass of neutrinos considered. This allows us to directly compare real-space predictions of the matter clustering ratio with real-space measures of the clustering ratio of any biased matter tracer,
i.e. ηg,R(r) ≡ ηR(r).
3.2 Redshift space
In redshift space, as described in Sec. 2, the apparent pos-ition of galaxies is modified according to the projection of their peculiar velocity along the line-of-sight. This effect dis-torts the clustering properties of the distribution and we thus expect its correlation function and variance to be af-fected. However, we expect redshift-space distortions not to affect the clustering ratio on linear scales (Eq. 17) as the effect is cancelling out into the ratio. In order to verify the accuracy of this approximation we created the redshift space catalogues of the FoFs and spherical overdensities in the simulations, moving the positions of the tracers along an arbitrary direction, chosen as the line-of-sight direction.
Having shown that the clustering ratio does not depend on the way we define the haloes nor on their mass tracer, from now on we focus on the dark matter halo catalogues identified with the Friend-of-Friends algorithm and we
com-pare the two extreme cases of Mν = 0 eV and Mν = 0.53
eV. Note that we use the Mν= 0 simulation as a reference
0.01 0.1 1 ξR (r ), r = 2R z = 0.48551 Bin 2 Bin 3 Bin 4 Bin 5 Bin 6 Bin 7 Bin 8 10 15 20 R [h−1Mpc] 1 10 p ξ h,R (r )/ξ R (r ) z = 1.05352 linear non-linear 10 15 20 R [h−1Mpc] FoF, Mν= 0.00 0.01 0.1 1 ξR (r ), r = 2R z = 0.48551 Bin 2 Bin 3 Bin 4 Bin 5 Bin 6 Bin 7 Bin 8 10 15 20 R [h−1Mpc] 1 10 p ξ h,R (r )/ξ R (r ) z = 1.05352 linear non-linear 10 15 20 R [h−1Mpc] FoF, Mν= 0.53
Figure 1. Correlation function of the Friend-of-Friends in the simulations with two different cosmologies. In the plot on the left there are the measurements in the ΛCDM cosmology, while in the plot on the right the ΛCDMν simulation with Mν= 0.53 eV. In both cases,
left panel is at redshift z = 0.48551 and right panel at z = 1.05352. In the top panel we show the correlation function measured in the mass bins presented in Tab. 2 (points) compared to the theoretical smoothed matter correlation function (black solid line). In the bottom panel there is the FoF-matter bias, computed as b =
q ξF oF
R (r)/ξR(r). The linear bias in the simulation with massive neutrinos
is larger than in the standard ΛCDM case, because, as neutrinos reduce clustering, massive haloes become rarer. We the bias values with a straight line between R = 16 and 22 h−1Mpc. As the fit shows, the linear bias is compatible with being constant with scale.
0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 ηR (nR ), R = 16 h − 1Mp c, n = 2 z = 0.48551 Mν= 0.00 eV Mν= 0.17 eV Mν= 0.30 eV Mν= 0.53 eV z = 1.05352 10 100 M [1012h−1M ] −10 0 10 % 10 100 M [1012h−1M ] FoF 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 ηR (nR ), R = 16 h − 1Mp c, n = 2 z = 0.48551 Mν= 0.00 eV Mν= 0.17 eV Mν= 0.30 eV Mν= 0.53 eV z = 1.05352 1 10 100 M [1012h−1M ] −10 0 10 % 1 10 100 M [1012h−1M ] SO
Figure 2. The clustering ratio of the Friend-of-Friends (FoF, left plot) and of the spherical overdensities with respect to the critical density (SO, right plot) computed in each mass bin at redshift z = 0.48551 (left panel of each plot) and z = 1.05352 (right panel) for all the neutrino masses. The smoothing radius is R = 16 h−1Mpc and the correlation length is twice the smoothing radius. Points are
measures in the simulations, while lines are the theoretical predictions. At z = 0.48551 the agreement between measures and predictions is better than 3% in the bins with masses < 30.43h−1M (i.e. where the bias is < 2), while in the bins with larger masses, corresponding
to larger values of bias, the agreement is at 5% level. Errors are larger in the highest mass bin because it is more sparsely populated. At z = 1.05352 the discrepancy between measures and predictions shows a specific dependence on the mass of the tracer. We consider this effect to be due a mass resolution effect in the simulation. Finally we point out that we find similar results for FoF and SO, meaning that the clustering ratio at these scales is insensitive to the linear bias irrespective of the mass tracer we choose.
properties are valid when neutrino are massless (see Bel & Marinoni 2014; Bel et al. 2014).
Fig. 3 shows, at the scales of interest, the independence of the clustering ratio from redshift-space distortions. The ratio between measurements of the clustering ratio in red-shift and in real space is of order 1, either with and without massive neutrino, both at z = 0.48551 and z = 1.05352.
In particular we show that we recover the results already obtained in other works (Bel & Marinoni 2012) in the ΛCDM case. In the case including massive neutrinos we find an agreement between redshift and real space measurements at 3% level at redshift z = 0.48551 and at better than 1%
on scales R & 16 h−1Mpc at z = 1.05352. In this case the
accuracy is higher at higher redshift as the growth of struc-tures is more linear.
Moreover we note that the agreement with predictions is better for the simulation with massive neutrinos with re-spect to the ΛCDM one. This is due to the fact that massive neutrinos lower the matter fluctuations (see the values of
σ8,cc in Tab. 1) and therefore tend to reduce the velocity
dispersion, resulting in redshift-space distortions that are more into the linear regime.
We therefore propose to use the clustering ratio as a cosmological probe to constrain the parameters of the cos-mological model. As a matter of fact, the analysis of the sim-ulations has shown that the clustering ratio, besides being independent from the matter tracer and the bias, is not af-fected by redshift-space distortions on linear scales. This
im-plies that Eq. (17), ηsg,R(r) ≡ ηR, still holds in the presence
of massive neutrinos, allowing us to directly compare cluster-ing ratio measurements in redshift-survey galaxy catalogues to the theoretical matter clustering ratio predictions.
4 RESULTS
4.1 Optimisation
In order to estimate and predict the clustering ratio of galax-ies, we need to choose two different scales: the smoothing scale R, i.e. the radius of the spheres we use for counting objects, and the correlation length r, that for simplicity we assume to be some multiple of the smoothing scale, r = nR. Choosing the best combinations of R and r is seminal to maximise the information we can extract from this statist-ical tool.
The smoothing scale R controls the scale under which we make our observable blind to perturbations. A sufficiently large value of R allows us to screen undesired nonlinear ef-fects, that would compromise the effectiveness of the clus-tering ratio. On the other hand, an excessively large smooth-ing scale can lead to more noisy measurements, since in the same volume we can accommodate fewer spheres. Moreover, if R is too large, the entire signal would be screened and the measurement would become of little interest.
Also for the correlation length, choosing small values of R and n implies coping with small-scale nonlinearities, which risk to invalidate the identity expressed in Eq. (17). Large values of correlation distances, however, would make it difficult to accommodate enough couples of spheres in the volume to guarantee statistical robustness. An addi-tional constraint comes from the strategy we adopt to fill
0.10 0.12 0.14 0.16 0.18 0.20 ηR (nR ), n = 2 z = 0.48551 real space, Mν= 0 eV redshift space, Mν= 0 eV real space, Mν= 0.53 eV redshift space, Mν= 0.53 eV z = 1.05352 10 15 20 R [h−1Mpc] 0.95 1.00 1.05 η s R(r )/η r R(r ) 10 15 20 R [h−1Mpc]
Figure 3. The clustering ratio smoothed on the scale R and at correlation length r = nR, n = 2 as a function of the smooth-ing scale. We show in red the measures in the ΛCDM simulation and in blue the ones in the simulation with the highest neutrino mass, Mν= 0.53 eV, which are the two extreme cases. Filled dots
are measures in real space, while empty dots in redshift space. In the bottom panel the ratio between the clustering ratio in red-hift space over the real space case is shown. Since on linear scales the monopole contribution coming from redshift-space distortions enhances the correlation function and the variance by the same multiplicative factor, we expect the clustering ratio to be unaf-fected. The ratio between redshift and real space measurements is, in fact, of order 1 with an accuracy better then 3%. This is even better confirmed in the case at redshift z = 1.055 (right) because at higher redshift the matter growth is more linear.
the volume with spheres and perform the count-in-cells. In this framework, if the correlation length is below twice the smoothing scales, r < 2R, the spheres of our motif of cells would overlap, resulting in an additional shot-noise contri-bution. For this reason, we only allow values of n > 2.
The main information we want to extract is the total neutrino mass. The sensitivity of the clustering ratio to this parameter can be quantified as an effective signal-to-noise ratio, defined as S/N =η ν R(r) − ηRΛ(r) σΛ R , (30)
where ηνR(r) is the clustering ratio measured in a simulation
with neutrino mass Mν, ηRΛ(r) is measured in the reference
ΛCDM simulation and σRΛis the uncertainty on the
cluster-ing ratio measured in the ΛCDM simulation. This quantity measures how much a massive neutrino cosmology is distin-guishable from a ΛCDM one, given the typical errors on the measures of the clustering ratio for the specific volume and number density of tracers, as a function of R and r. Fig. 4 shows the (n, R) plane, constructed as a grid with correla-tion lengths n ∈ [2, 2.75] with step ∆n = 0.05 and smooth-ing scales R ∈ [15, 30] with step ∆R = 1, all distances besmooth-ing
expressed in units of h−1 Mpc. At each point on the grid
a color is associated, representing the value of this effect-ive signal-to-noise ratio. The effect of masseffect-ive neutrinos is, as expected, appreciable on small scales (both small n and small R), and eventually becomes negligible moving towards large scales.
2.0 2.2 2.4 2.6 n 16 18 20 22 24 26 28 R [h − 1Mp c] Mν= 0.17 eV, z = 0.48551 0.5000 1.0000 1.4518 [ην R(r)− ηRΛ(r)]/σΛη
Figure 4. The effect of neutrinos on the clustering ratio com-pared to a ΛCDM cosmology. In the (n, R) plane, we plot colour contours corresponding to (ηR(r, ν) − ηR(r, ΛCDM)/ση(ΛCDM).
As expected, the sensitivity to the neutrino total mass increases at small smoothing scales and correlation lengths (red regions).
While we want to maximise the effects of neutrinos, we want to minimise errors. In particular, we can define a theor-etical error, that accounts for the combinations of smoothing radii R and correlation lengths r where the assumptions un-der which we can apply the identity expressed in Eq. (17) break down. Such theoretical error can be quantified as
δth=
ηR(r) − ηRth(r)
ση
, (31)
where ηR(r) is the clustering ratio measured in the
simula-tion with a given cosmology, ηRth(r) is the prediction obtained
with a Boltzmann code, and σηthe uncertainty on the
meas-urement. In Fig. 5 we show as a color map the values that we obtain for this theoretical error in the same (n, R) plan introduced above. We can see that on very small scales the effect of nonlinearities is not negligible, and we cannot use the clustering ratio as an unbiased observable.
From Fig.s 4 and 5 we see that we need to balance between the requirement coming from the signal-to-noise ratio (that is maximum on small scales), and those from the theoretical errors (that is minimum on large scales). We introduce, therefore, a way of combining these pieces of in-formation into a single colour map, which we use to seek the sweet-spots, in this parameter space, where both conditions are satisfied.
First, we define a combined percentage error (that ac-counts both for statistical errors and discrepancies from the model) as δcombined= ( ηνR(r) − η ν,th R (r) ην R(r) − ηΛR(r) ) ( σΛη ην R(r) − ηΛR(r) ) . (32)
The quantity in the first parenthesis is related to how much the statistical error is important with respect to the effects of neutrinos, while the second parenthesis is a weight that accounts for the typical uncertainty on the measurement in each bin of n and R. Finally, we define the neutrino contrast
2.0 2.2 2.4 2.6 n 16 18 20 22 24 26 28 R [h − 1Mp c] Mν= 0.17 eV, z = 0.48551 0.500 1.000 1.500 2.000 3.508 [ην R(r)− ηRth(r)]/σην
Figure 5. Discrepancy between the clustering ratio measured in the simulation and the theoretical prediction in the (n, R) plane. Colours represent the quantity (ηR(r) − ηRth(r))/σ
(
Rr), the blue
regions being the ones with the best agreement with the pre-dictions. Sufficiently large smoothing scales screen the effects of the nonlinear growth of perturbations, allowing us to exploit the clustering ratio as a cosmological probe. By smoothing our distri-bution on scales R > 19 h−1 Mpc we ensure an agreement with
the model better then ∼ 1.5 standard deviations.
as C = S/N max(S/N) − δcombined max(δcombined) . (33)
Here we have normalised the signal/noise defined in Eq. (30) and the combined error defined in Eq. (32) to their respect-ive maxima (on the considered grid) and we are interested in finding the regions where this contrast is dominated by the signal/noise, i.e. where C(n, R) ∼ 1. In Fig. 6 we show the neutrino contrast on the (n, R) grid, for the simulation
with Mν= 0.17 eV at redshift z = 0.48551.
We have repeated this analysis for the three massive
neutrino simulations (with Mν = 0.17, 0.30, 0.53 eV) and
for different redshifts, spanning the range from z = 0.48551 to z = 2.05053. Our conclusion is that the combination R =
22 h−1Mpc, n = 2.1 is the most viable candidate for all
these cosmologies and redshifts.
4.2 Likelihood
We aim at comparing measurements and predictions of the clustering ratio, in order to find the set of parameters of the model that maximizes the likelihood. We exploit the different dependence of measurements and predictions on the cosmological model. In particular, the measured value of the clustering ratio depends on the way we convert redshifts and angles into comoving distances, that depends on the
total matter density Ωm, on the dark energy fraction ΩΛ
and on the background expansion rate H(z).
On the other hand, the theoretical prediction for the clustering ratio depends on the entire cosmological model and is therefore sensitive also to the value of the total
neut-rino mass Mν.
2.0 2.2 2.4 2.6 n 16 18 20 22 24 26 28 R [h − 1Mp c] Mν= 0.17 eV, z = 0.48551 −0.526 0.000 0.500 0.786 Neutrino Contrast
Figure 6. In this colour plot we subtract to the neutrino signal-to-noise normalised to 1 a combined error normalised in the same fashion. Details on the definition are in the text. We are interested in the regions in the (n, R) plane where the neut-rino signal-to-noise dominates (∼ 1) on the error (∼ 0), graph-ically visible as hot spots. The region with smoothing scales 20 < R < 23 h−1Mpc
seems to be the most promising. In particular, by repeating this test for different redshifts and neutrino masses, we chose as our
candidate scales R = 22 h−1Mpc, n = 2.1.
namely the baryon and cold dark matter density parameters
Ωbh2 and Ωcdmh2, the Hubble parameter H0, the optical
depth at the recombination epoch τ , the amplitude of the
scalar power spectrum at the pivotal scale Asand the scalar
spectral index ns. Moreover, we extend this parametrization
with two additional free parameters, the total neutrino mass
Mνand the equation of state of the dark energy fluid w. The
most general vector of parameters therefore is
p = {Ωbh2, Ωcdmh2, H0, τ, As, ns, Mν, w}.
We follow Bel & Marinoni (2014), who showed that the like-lihood function of the clustering ratio (given a fixed set of parameters) is compatible with being a Gaussian. Therefore
we will compute the logarithmic likelihood as ln L = −χ2/2
(apart from a normalization term) where
χ2(p) =X i (ηR,i(r) − ηR,ith (r))2 σ2 ηi , (34)
where we neglect the covariance between the different red-shift bins.
We account for the dependence of the measurements on the cosmological model assumed, induced by the cosmology-dependant conversion of redshifts into distances, whenever we compare measurements of the clustering ratio (obtained in the fiducial cosmology) to its predictions (in a generic cosmology). That is, when computing the likelihood for the set of parameters ϑ, we must keep in mind that the measured value has been computed in a different cosmology, the one
with the fiducial set of parameters ϑF.
We keep the measurements fixed in the fiducial cosmo-logy and rescale the predictions accordingly. We consider
that, due to Alcock-Paczy´nski effect, at same redshift and
angular apertures we can associate different lengths
depend-ing on cosmology (Alcock & Paczy´nski 1979).
Our measurements depend on distances only through the smoothing scale R. This is because the correlation length is always expressed as a multiple of the smoothing scale, r = nR. This means that, since the measure has been obtained in the fiducial cosmology using spheres of radius
RF= 22 h−1
Mpc, they need to be compared to predictions obtained in a generic cosmology using a smoothing length
R = αRF, where α is our Alcock-Paczy´nski correction.
We write the Alcock-Paczy´nski correction α as
(Eisen-stein et al. 2005) α = " EF(z) E(z) DA DF A 2#1/3 , (35)
where E(z) ≡ H(z)/H0 is the normalised Hubble function
and DA the angular diameter distance. Therefore, we are
going to compare
ηF,sg,R(nR) ≡ ηαR(nαR), (36)
the left hand side of Eq. (36) being the clustering ratio of galaxies measured in redshift space assuming the fiducial cosmology, while the right hand side is the predicted clus-tering ratio for matter in real space, rescaled to the fiducial cosmology to make it comparable with observations.
In order to efficiently explore the parameter space we have modified the public code CosmoMC (Lewis & Bridle 2002), adding a likelihood function that implements this pro-cedure.
4.3 Constraints using SDSS data
We measure the clustering ratio in the 7th(Abazajian et al.
2009) and 12th (Alam et al. 2015) data release of the Sloan
Digital Sky Survey (SDSS) by smoothing the galaxy dis-tribution with spherical cells of radius R and counting the objects falling in each cell. We divide the sample into three
redshift bins that have mean redshifts ¯z = {0.29, 0.42, 0.60}.
The first redshift bin is extracted from the DR7 catalogue, while the two bins at higher redshift come from the DR12 catalogue, after removing the objects already present in the other bin.
To perform the count-in-cell procedure, we convert
red-shifts into distances, assuming a cosmology with H0 =
67 km s−1Mpc−1, Ωm= 0.32 and in which we fix the
geo-metry of the universe to be flat, Ωk = 0, forcing ΩΛ =
1 − Ωr− Ωm. Therefore, this is to be considered our
fidu-cial cosmology. We compute the clustering ratio using the estimators presented in Sec. 2.1, employing our optimised
smoothing size and correlation length, R = 22 h−1Mpc and
r = 2.1 R. Our measures of the clustering ratio in each red-shift bin are
at 0.15 6 z 6 0.43, ηg,R(r) = 0.0945 ± 0.0067,
at 0.30 6 z 6 0.53, ηg,R(r) = 0.0914 ± 0.0055,
at 0.53 6 z 6 0.67, ηg,R(r) = 0.1070 ± 0.0110.
Details on the computation of the clustering ratio and its er-rors in the SDSS catalogue can be found in Bel et al. (2015), where, though, measures are performed assuming a different fiducial cosmology.
0.952 0.960 0.968 0.976 ns 0.117 0.120 0.123 Ωc h 2 0.25 0.50 0.75 1.00 Σ mν 60 63 66 69 H0 0.952 0.960 0.968 0.976 ns 0.117 0.120 0.123 Ωch2 0.25 0.50 0.75 1.00 Σmν CMB CMB + CR
Figure 7. Joint posterior distribution obtained using Planck tem-perature and polarization data and the clustering ratio measured in SDSS DR7 and 12. We fit a cosmological model with seven free parameters, the six baseline parameters of Planck and the neutrino total mass, but here only four of them are shown.
In Fig. 7 we show, for some relevant parameters, the joint posterior distribution obtained fitting at the same time the Planck temperature and polarization data and the clus-tering ratio measurements in SDSS DR7 and DR12, leaving free to vary the six baseline parameters and the total
neut-rino mass Mν. Already by eye, adding the clustering ratio
to the CMB information does not seem to improve much the upper bound of the total neutrino mass parameter. In general, the most significant improvement seems to occur on the constraint of the cold dark matter density parameter.
Moreover, we have also checked how constraints change when we leave the equation of state of dark energy, w, as an additional free parameter. As a matter of fact, w is known to be strongly degenerate with the other parameters of the model, when only CMB data are used. In general, we need information from a geometrical probe sensitive to the late time universe in order not to find non-physical solutions. Fig. 8 shows that the clustering ratio is indeed able to break such degeneracy.
Also the combination of other cosmological probes can help breaking degeneracies and tightening constraints. For this reason we compare the constraining power of the clus-tering ratio to that of two other observables, the fit of the BAO peak in the correlation function measured by the BOSS collaboration in the DR11 CMASS and LOWZ datasets (An-derson et al. 2014) and the lensing of the CMB signal due to the intervening matter distribution between the last scatter-ing surface and us, where the amplitude of the lensscatter-ing
po-tential, AL, has been kept fixed to 1 (Planck Collaboration
et al. 2016).
In the first part of Tab. 4 we show the mean, 68% and 95% levels obtained for the different parameters combining the likelihoods presented above. To better show the
beha-viour of the clustering ratio with respect to the other probes considered, in Fig.s 9-10, we focus particularly on the
para-meters w, Mν and H0.
In general, adding the clustering ratio considerably im-proves on the parameter constraints obtained with CMB data alone, especially when also w is free to vary. In par-ticular, the clustering ratio is able to break the degeneracy between w and the other cosmological parameters, that af-fects the constraints drawn with the sole CMB data. On the other hand, the clustering ratio does not seem to improve
much the constraint on the Mνparameter, especially when
compared to probes such as the CMB lensing and the BAO peak position.
The clustering ratio proves to be extremely sensitive to
the cold dark matter fraction Ωcdmh2, as adding the
cluster-ing ratio to the CMB analysis results in a 12% improvement on the 95% confidence level.
To improve our understanding of the results presented in the previous section, we investigate how well the cluster-ing ratio allows us to recover a certain known cosmology.
To this purpose, we use the measurements of the clus-tering ratio in one of the DEMNUni simulations, the one
with Mν = 0.17 eV, which represents the closest value to
the current available constraints on the neutrino total mass. The clustering ratio is measured in the simulation at the same redshifts, and with the same binning, as in the SDSS data. The error on each measurement in the simulation is taken to be the one obtained from the SDSS measurements. The likelihood using the CMB data is computed in this case fixing the bestfits to the values of the parameters in the cosmology of the simulation, and employing the covari-ance matrix contained in the publicly available Planck data release.
The posterior distribution obtained with this procedure is shown in Fig. 11, while the second part of Tab. 4 summar-izes the improvements on the constraints on the parameters that we obtain adding the clustering ratio. We correctly re-cover the bestfits of our known cosmology, with errors com-parable with the true ones. We conclude that the reason why we did not achieve a significant improvement on the constraint on the total neutrino mass using the SDSS data resides in the fact that the clustering ratio requires smaller error bars to be effective in constraining such parameter. We can therefore expect that, with upcoming, large galaxy redshift surveys, the clustering ratio will reach a larger con-straining power.
We test such an hypothesis in the next section, ana-lysing the clustering ratio expected for a Euclid-like galaxy redshift survey, in combination with CMB data.
4.4 Forecasts for a Euclid-like galaxy redshift
survey
In order to forecast the constraining power of the cluster-ing ratio, expected from a future, Euclid-like galaxy redshift survey, we construct the synthetic clustering ratio data in the following way:
• We imagine to have 14 redshift bins, from z = 0.7 to z = 2, with ∆z = 0.1
• In each redshift bin, the synthetic measurement of the clustering ratio is given by the predicted clustering ratio
0.952 0.960 0.968 0.976 ns 0.0220 0.0224 0.0228 Ωb h 2 0.114 0.117 0.120 0.123 Ωc h 2 −1.5 −1.2 −0.9 −0.6 w 0.03 0.06 0.09 0.12 τ 3.00 3.06 3.12 3.18 ln(10 10 As ) 60 70 80 90 H0 0.952 0.960 0.968 0.976 ns 0.0220 0.0224 0.0228 Ωbh2 0.114 0.117 0.120 0.123 Ωch2 −1.5 −1.2 −0.9 −0.6 w 0.03 0.06 0.09 0.12 τ 3.00 3.06 3.12 3.18 ln(1010A s) CMB CMB + CR
Figure 8. Joint posterior distribution obtained employing CMB temperature and polarization data from Planck and the clustering ratio measurements from SDSS DR7 and DR12 catalogues. Besides the six standard parameters of the model, also the equation of state of dark energy is left free.
(computed using a Boltzmann code), to which we add a small random noise (within 1 standard deviation).
• We measure the errors (at the same redshifts) in the DEMNUni simulations; the errors in the simulations are then rescaled, according to the operative formula presen-ted below, to match the volume and number density of our Euclid-like survey.
The relative error on the clustering ratio depends on the volume and number density of the sample, and can be
parametrised, following Bel et al. (2015), as δη η = AV −1/2 exp 0.14 ln ρ − ln 2 ρ 2 ln(0.02) (37)
where V is the volume expressed in h−3Mpc3, ρ is the object
number density in h3Mpc−3and A is a normalization factor
computed with the reference volume and number density. We use these data to explore the posterior distribution of the parameters of the model. The results are shown in Fig. 12, and the constraints are shown in the last part of Tab. 4.
−2.8 −2.4 −2.0 −1.6 −1.2 −0.8 w 50 60 70 80 90 100 H0 CMB CMB + CR CMB + BAO CMB + CR + BAO −2.8 −2.4 −2.0 −1.6 −1.2 −0.8 w 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Σ mν CMB CMB + CR CMB + BAO CMB + CR + BAO
Figure 9. Degeneracy between the equation of state of dark energy, w, and, on the left, the Hubble parameter today H0, and on the
right the total neutrino mass Mν. The considered likelihoods are the one computed using Planck data alone, and its combinations with
the clustering ratio measured in SDSS DR7 and 12, the BAO position from SDSS DR11 or both.
−2.8 −2.4 −2.0 −1.6 −1.2 −0.8 w 50 60 70 80 90 100 H0 CMB CMB + CLens CMB + CR CMB + CR + CLens −2.8 −2.4 −2.0 −1.6 −1.2 −0.8 w 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Σ mν CMB CMB + CLens CMB + CR + CLens CMB + CR
Figure 10. Degeneracy between the equation of state of dark energy, w, and, on the left, the Hubble parameter today H0, and on the
right the total neutrino mass Mν. The considered likelihoods are the one computed using Planck data alone, and its combinations with
the clustering ratio measured in SDSS DR7 and 12, the lensing of the CMB or both.
In this case there is a much larger improvement on the constraints of all the parameters. The neutrino total mass parameter improves by 14% on the 95% limit with respect to using Planck alone. Most notably, the constraint on the cold
dark matter density parameter, Ωcdmh2, improves by over
40%. Also the spectral index ns shows a 10% improvement
and the constraint on the Hubble constant H0 improves by
20%.
This means that, when new data, covering a larger volume, will be available, clustering ratio measurements are expected to contribute with a significant improvement on the constraints on the parameters of the cosmological model.
We also note that, as more different observations are carried out, it becomes very interesting to enhance the constraining power of the clustering ratio also combining its measurements in different datasets. This can be easily done since the clustering ratio is a single measurement, thus scarcely dependent on the survey geometry.
5 SUMMARY AND CONCLUSION
Neutrino effects are being increasingly included in cosmolo-gical investigations, becoming in fact part of the standard cosmological model. Thanks to these investigations, the de-scription of the statistical properties of the universe is gain-ing the precision required by forthcomgain-ing experiments and, at the same time, neutrino physics gains tighter constraints. In this work we have considered the clustering ratio, an observable defined as the ratio between the smoothed correlation function and variance of a distribution, and ex-tended its range of applicability to cosmologies that include a massive neutrino component. As a matter of fact, the clus-tering ratio, which has already been tested in ΛCDM cos-mologies including only massless neutrinos, is unbiased and independent from redshift-space distortions on linear scales. As massive neutrinos introduce characteristic scale depend-encies in the clustering of galaxies (and matter), such pecu-liar properties of the clustering ratio needed to be confirmed (or denied) in this cosmological framework.
We divided our analysis into two steps: first, we stud-ied the properties of the clustering ratio in simulations that
0.95 0.96 0.97 ns 0.1150 0.1175 0.1200 0.1225 Ωc h 2 0.15 0.30 0.45 0.60 P mν 62.5 65.0 67.5 70.0 H0 0.95 0.96 0.97 ns 0.11500.11750.12000.1225 Ωch2 0.15 0.30 0.45 0.60 Pm ν CMB CMB + CR
Figure 11. Joint posterior distribution obtained using Planck temperature and polarization data and the clustering ratio. Be-stfits here are fixed, errors for Planck come from the publicly available covariance matrix, errors on the clustering ratio have been compute in SDSS DR7 and DR12. Four of the seven free parameters are shown.
include massive neutrinos; afterwards, we used the cluster-ing ratio to compute the likelihood of the parameters of the cosmological model, using both real data and forecasts of future data.
In the first part of this work, we employed the DEM-NUni simulations to test the clustering ratio in the pres-ence of massive neutrinos. These are the largest available simulations that include massive neutrinos as a separate particle species along with cold dark matter. We computed the clustering ratio using different tracers (dark matter FoF haloes and spherical overdensities), divided into
differ-ent mass bins (spanning the interval from ∼ 6 × 1011 to
& 1014 h−1
M ), and we explore different choices of
neut-rino mass (Mν= {0, 0.17, 0.3, 0.53} eV) in real and redshift
space.
From such analysis we conclude that the properties of the clustering ratio hold also in cosmologies with massive neutrinos. In particular its main property, the fact that the galaxy clustering ratio in redshift space is directly compar-able to the clustering ratio predicted for matter in real space on a range of linear scales, is proven valid.
We have therefore moved to employing the clustering ratio as a cosmological probe to find the set of parameters of the model that maximizes the likelihood function, given a set of data. We have used the data from the SDSS DR7+DR12 catalogue. We have computed the clustering ratio in three redshift bins and used these measures in combination with the temperature and polarization anisotropies of the CMB measured by the Planck satellite to explore the likelihood in parameter space with an MCMC approach. We find that the clustering ratio is able to break the degeneracy, present in the CMB data alone, between the equation of state of dark
matter, w, and the other parameters. Moreover it improves the 95% limit on the CDM density parameter by ∼ 12%. However, we do not find an appreciable improvement in the constraint on the neutrino total mass.
By analysing simulations we conclude that we blame such lack of improvement on the statistical errors, which, with current data, are not yet competitive enough. We have therefore tested the constraining power of the clustering ra-tio using the error bars expected from a Euclid-like galaxy survey.
In this case we find that not only does the clustering ra-tio greatly improve (with respect to using CMB data alone) the constraint on the CDM density parameter (shrinking
the 95% limit by ∼ 40%) and on the Hubble parameter H0
(whose 95% limit improves by 20%), but it is also able to im-prove by ∼ 14% the 95% upper bound on the total neutrino mass.
In conclusion, the clustering ratio appears to be a valu-able probe to constrain the parameters of the cosmological model, especially with upcoming large galaxy redshift sur-veys. Being easy to model and to measure, it provides us with a powerful tool to complement other approaches to galaxy clustering analysis, such as the measurement of the galaxy correlation function or power spectrum. Moreover, we note that, given the simplicity of combining the cluster-ing ratio measured in different surveys, we expect its true constraining power to emerge when it will be measured in a number of different datasets.
0.95 0.96 0.97 ns 0.0220 0.0224 0.0228 Ωb h 2 0.114 0.117 0.120 0.123 Ωc h 2 0.15 0.30 0.45 0.60 Σ mν 0.04 0.08 0.12 0.16 τ 2.96 3.04 3.12 ln(10 10 As ) 62.5 65.0 67.5 70.0 H0 0.95 0.96 0.97 ns 0.0220 0.0224 0.0228 Ωbh2 0.114 0.117 0.120 0.123 Ωch2 0.15 0.30 0.45 0.60 Σmν 0.04 0.08 0.12 0.16 τ 2.96 3.04 3.12 ln(1010A s) CMB CMB + CR
Figure 12. Joint posterior distribution obtained combining Planck temperature and polarization data and the clustering ratio measured in a Euclid-like galaxy survey. For CMB data, errors come from the publicly available covariance matrix, for the clustering ratio errors have been measured in the DEMNUni simulations and rescaled to match the volume and number density of our mock survey.
M.
Zennar
o,
J.
Bel,
J.
Dossett,
C.
Carb
one,
L.
Guzzo
Ωbh2 Ωch2 τ Mν Pl (fixed w) 0.02222 ±0.00017 ±0.00033 0.11978 ±0.00147 ±0.00291 0.07851 ±0.01713 ±0.03355 0.16722 < 0.19150 < 0.49402 Pl + CR (fixed w) 0.02222 ±0.00016 ±0.00031 0.11972 ±0.00128 ±0.00255 0.07801 ±0.01744 ±0.03330 0.15795 < 0.18088 < 0.47835 Pl 0.02222 ±0.00016 ±0.00034 0.11971 ±0.00142 ±0.00281 0.07737 ±0.01793 ±0.03483 0.22153 < 0.26698 < 0.60851 Pl + CR 0.02216 ±0.00017 ±0.00033 0.12042 ±0.00149 ±0.00290 0.07469 ±0.01754 ±0.03403 0.20304 < 0.24510 < 0.53081 Pl + CLens 0.02217 ±0.00017 ±0.00035 0.11967 ±0.00153 ±0.00299 0.06927 ±0.01749 ±0.03420 0.32882 ±0.19711 < 0.67219 Pl + BAO 0.02225 ±0.00015 ±0.00030 0.11949 ±0.00134 ±0.00263 0.07728 ±0.01705 ±0.03304 0.11571 < 0.14235 < 0.30423 Pl + CLens + CR 0.02213 ±0.00016 ±0.00032 0.12026 ±0.00145 ±0.00288 0.06909 ±0.01681 ±0.03237 0.29440 ±0.17524 < 0.59471 Pl + BAO +CR 0.02223 ±0.00015 ±0.00030 0.11965 ±0.00132 ±0.00261 0.07731 ±0.01666 ±0.03147 0.10844 < 0.13143 < 0.28339 w ln(1010A s) ns H0 Pl (fixed w) -1.00000 − − 3.09167 ±0.03332 ±0.06510 0.96531 ±0.00478 ±0.00951 66.36205 +1.93320−0.79827 ±3.14533 Pl + CR (fixed w) -1.00000 − − 3.09042 ±0.03375 ±0.06497 0.96550 ±0.00456 ±0.00906 66.47015 +1.72753−0.68157 ±2.93440 Pl -1.68615 ±0.29543 ±0.59285 3.08881 ±0.03490 ±0.06751 0.96500 ±0.00473 ±0.00953 86.91446 +12.41268−4.70920 ±15.59971 Pl + CR -1.25376 ±0.24254 ±0.52000 3.08537 ±0.03368 ±0.06550 0.96389 ±0.00487 ±0.00967 73.04263 ±6.83317 ±14.51362 Pl + CLens -1.67628 ±0.35984 ±0.66750 3.07148 ±0.03387 ±0.06578 0.96452 ±0.00502 ±0.00978 84.62450 +14.64438−5.54025 ±16.85512 Pl + BAO -1.05867 ±0.07959 ±0.16354 3.08846 ±0.03305 ±0.06428 0.96612 ±0.00451 ±0.00886 68.60048 ±1.67361 ±3.36935 Pl + CLens + CR -1.22298 ±0.23981 ±0.51234 3.07259 ±0.03189 ±0.06176 0.96385 ±0.00493 ±0.00946 71.08454 ±6.31034 ±13.48706 Pl + BAO + CR -1.05267 ±0.07749 ±0.15731 3.08903 ±0.03209 ±0.06098 0.96595 ±0.00453 ±0.00888 68.42143 ±1.63637 ±3.26416 Ωbh2 Ωch2 τ Mν Pl (fixed w) 0.02244 ±0.00016 ±0.00034 0.11948 ±0.00157 ±0.00314 0.09332 ±0.01644 ±0.03291 0.20477 < 0.26441 < 0.42993 Pl + CR (fixed w) 0.02242 ±0.00015 ±0.00030 0.11950 ±0.00136 ±0.00265 0.09199 ±0.01659 ±0.03229 0.20843 ±0.11909 < 0.41013 w ln(1010A s) ns H0 Pl (fixed w) -1.00000 − − 3.05893 ±0.03169 ±0.06381 0.95937 ±0.00518 ±0.01015 66.65175 ±1.43657 ±2.67394 Pl + CR (fixed w) -1.00000 − − 3.05620 ±0.03264 ±0.06422 0.95969 ±0.00458 ±0.00918 66.60347 ±1.23252 ±2.41851 Ωbh2 Ωch2 τ Mν Pl (fixed w) 0.02243 ±0.00016 ±0.00031 0.11942 ±0.00146 ±0.00290 0.09335 ±0.01747 ±0.03445 0.20513 ±0.12382 < 0.43157 Pl + CR (fixed w) 0.02245 ±0.00013 ±0.00026 0.11926 ±0.00084 ±0.00167 0.09422 ±0.01655 ±0.03298 0.17749 ±0.11133 < 0.37693 w ln(1010A s) ns H0 Pl (fixed w) -1.00000 − − 3.05859 ±0.03378 ±0.06670 0.95983 ±0.00491 ±0.00964 66.66588 ±1.38464 ±2.62998 Pl + CR (fixed w) -1.00000 − − 3.06008 ±0.03262 ±0.06476 0.96046 ±0.00443 ±0.00854 66.98261 ±1.12263 ±2.18365Table 4. Mean, 68% and 95% levels of the marginalised posterior distributions. In the first part of the table the datasets of Planck’s CMB temperature and polarization anisotropies, BOSS measurement of the BAO peak, and Planck’s CMB lensing signal are used in combination with the clustering ratio measured in the SDSS DR7+12 sample. In the second part of the table bestfits have been fixed to the ones of the fiducial cosmology, errors for Planck are obtained from the publicly available Planck parameter covariance matrix and error for the clustering ratio are the ones measured in SDSS DR7 and DR12. Finally, in the last part of the table, bestfits have been fixed to the ones of the fiducial cosmology, errors for Planck are obtained from the publicly available Planck parameter covariance matrix and errors for the clustering ratio are the ones predicted for a Euclid-like galaxy survey, following the procedure described in the text.
c 0000 RAS, MNRAS 000 , 000–000
